Ai intergration

Feature request
1

Can they be a way to Intergrate AI abilities into the app. Like using AI for summary and transcript. This will be awesome.
Thanks for everything you guys have already done.

1 Like
2

Doing that would require too much storage if done locally or hosting a server which is expensive or using a 3rd party service which is not private.

Try to install whisper yourself and you’ll see how much more storage and power would be needed to get basic transcripts to work

3

Android (13 at least) had a built in transcript tool called ‘live transcription’. It’s the square with dotted lines across it and a diagonal line though it when off below the vol gauge after using the vol buttons, it might need to be enabled tn the accessible settings and it is quite heavy on battery

1 Like
4

Hey. I run a podcast data platform that transcribes podcasts at scale, and would be happy to talk about providing transcripts to the AntennaPod.

I know AntennaPod it’s a free app so also happy to try to find a way to do it for free.

If any of the AntennaPod would like to talk about this then feel free to DM me.

5

@Joe_Tannorella Thank you for engaging on this. For us to move forward, I think we should get into a private chat/video with @ByteHamster , @keunes and @tonytamsf to see how this could work, whether there is mutual alignment of interests, and how it can benefit the ecosystem?

Providing this for free is awesome, I suspect it gets complicated when AP has to depend on another proprietary service.

2 Likes
6

fwiw, FUTO voice input is a foss app which has a local whisper model for speech-to-text. it works very well and fast on my Galaxy A55 (1-2y.o., 350eur) and also worked on my previous phone, Galaxy M21 (4+ y.o., 200eur, albeit slower).

it’s definitely a large and complex feature, but i think it could be done in a local, privacy-respecting way and would be awesome, as spotify has also introduced AI transcripts recently. It would also be very convenient to be able to full-text search across transcripts of all episodes of a podcast.

I imagine this could be implemented in two steps:

  1. each user can enable the transcriptions and have it live-transcribe or transcribe in advance for selected episodes. the transcript could be edited or re-synced. then, it’s stored in the antennapod db.

  2. a central, optional service where each antennapod user would, upon opting in, share their transcriptions with all other antennapod users, and also download any available transcriptions for episodes they’re going to listen to. obviously this wouldn’t be free to host & run, but it’s basically just sharing plaintext, so it’s much cheaper than doing the ai transcriptions in the cloud.